Datasets:
| language: | |
| - en | |
| license: cc-by-4.0 | |
| task_categories: | |
| - text-to-speech | |
| - audio-to-audio | |
| pretty_name: Dataset with DAC Codes | |
| tags: | |
| - audio | |
| - speech | |
| - dac | |
| - codec | |
| - s2s | |
| - dia | |
| # Dataset with DAC Codes | |
| This dataset adds DAC codec codes to [parler-tts/libritts_r_filtered](https://huggingface.co/datasets/parler-tts/libritts_r_filtered). | |
| ## Dataset Description | |
| Each sample contains: | |
| - **audio**: Audio resampled to 44.1kHz (DAC's native rate) | |
| - **codes**: 9-layer DAC codec codes (list of 9 lists of integers, vocab 0-1027) | |
| - **text**: Text transcription (from `text_normalized` column) | |
| ## Stats | |
| - **Source**: parler-tts/libritts_r_filtered | |
| - **Splits**: train.clean.100 | |
| - **Samples**: 32,215 | |
| - **Audio Sample Rate**: 44.1kHz | |
| - **Codec**: DAC (descript-audio-codec) with 9 codebooks, vocab size 1028 | |
| ## Usage | |
| ```python | |
| from datasets import load_dataset | |
| ds = load_dataset("mazesmazes/libritts-dac", split="train") | |
| sample = ds[0] | |
| codes = sample["codes"] # 9 lists of codec indices | |
| text = sample["text"] | |
| ``` | |
| ## License | |
| Same as source dataset. | |